9

Both issues are hard to measure statistically because very few will pay attention to it, and if it ever happens, the persons involved will quickly correct it and try to hide it from others.

However, there is technology to measure reliability (reproducibility and distortion) of barcode reading, such as this.

Are there any studies comparing the relative frequency of barcode misreads and cashiers giving out the wrong change?

  • 6
    Barcodes encode product-id not price tag. Usually if a barcode is misread then a different product appears on the reading machine. From my experience, the only time I saw a misread barcode the machine showed **93 tones** of cheese. – Alexandru May 23 '11 at 14:43
  • 1
    @Alexandru: in many cases Barcode also encodes the quantity of the item (especially items sold in weight); but you're right, in most cases misread barcode would give a totally off data that people would likely immediately notice. – Lie Ryan May 23 '11 at 15:10
  • 3
    Note that the most common barcodes include some sort of checksum (e.g. [EAN](http://en.wikipedia.org/wiki/European_Article_Number) has a check digit), so a misread would in most cases produce an invalid code. – Piskvor left the building May 23 '11 at 17:21
  • 2
    Sacnning errors must come in at least three sorts: (1) mis-scan, but as Piskvor says there is some protection against that (2) mis-labeling (wrong bar code on the product) which should be rare with commercial products but has been used to steel things (by covering the barcode with that for a much cheaper item) and (3) data-entry error in the database that backs up the scanner (which is probably the case with Alexandru's story). – dmckee --- ex-moderator kitten May 23 '11 at 17:39
  • @Lie Ryan: At the grocery stores I frequent, prices are either per item or per unit weight. Individual items have their bar codes, and the bar code reader doubles as a scale (I don't know how accurate it is), so in my experience here in the Midwest US the barcode doesn't encode quantity. – David Thornley May 24 '11 at 23:46
  • 4
    Due to the check digit, a barcode would have to be *very badly* misread to produce a number (most misreads simply produce an error output). Further, that number has to correspond to a real item in the database, and UPC codes in use today for products are not fully packed - you can't guarantee that by changing one number to another, you'll get a valid product, nevermind a valid product that this store actually carries and has a database entry for. All these factors combined mean that there's a very, very, very tiny chance of this happening. – Adam Davis Aug 08 '11 at 19:53
  • @AdamDavis: With some barcode formats such as Interleaved 2 of 5, a scan which enters and/or exits the top or bottom of the barcode may be misinterpreted as a legitimate barcode which encodes a different amount of data, possibly in a completely different format. – supercat Jun 15 '15 at 19:18
  • @supercat While the possibility exists theoretically, most sane systems include start and end codes which, if not encountered, will not interpret it as a valid code. Further, the check code would add significant protection from such misreads as well. So while a simple barcode without start and end codes could be read in fragments, and possibly even with a check digit one out of 10 times for a single digit check digit UPC and all other "sane" complete barcode systems would not have this problem - you would not consider a fragment of a UPC code as valid for these and other reasons. – Adam Davis Jun 15 '15 at 19:40
  • @AdamDavis: With interleaved 2 of 5, the start and end sequences are both very short, and appear very frequently within the barcode, so an i2of5 code truncated on either end would have a significant likelihood of reading as a shorter i2of5 code. UPC uses start/mid/stop code which cannot appear within a valid data stream, so the problem doesn't exist with that format, but many store scanners will accept formats in addition to UPC. – supercat Jun 15 '15 at 20:00
  • @supercat *"i2of5 code truncated on either end would have a significant likelihood of reading as a shorter i2of5 code."* This is an interesting assertion to make. Can you create an example code that contains a completely valid, different, code inside it as a fragment? If it's a significant likelihood, then it should be fairly trivial to force the situation. – Adam Davis Jun 15 '15 at 20:06
  • A pair of digits 35, 36, 55, 56, 65, or 66 will have an encoding which ends wide-space, narrow-bar, narrow-space, narrow-bar, narrow-space. If the scanning beam enters in that wide space, the remaining encoding will be a valid i2of5 code, save only for the check digit (which would have a one in ten chance of being correct). – supercat Jun 15 '15 at 20:17

1 Answers1

7

The question is confusing two separate issues:

Cashiers can give wrong change independently of whether bar-codes or price-tags are used.

Bar-codes are intended to lead to fewer data-entry errors and faster data-entry than price-tags.

Here is a quote from the Bar-Code FAQ at adams1.com.

What are the advantages and disadvantages of barcode? There are two basic advantages to barcode over manual data entry: Speed, and Accuracy. For 12 characters of data, keyboard entry takes 6 seconds. Scanning a 12 character barcode takes .3 seconds. The error rate for typing is one substitution error in every 300 characters types. Error rated for barcode range from 1 substitution error in every 15,000 to 36 trillion characters scanned (depending on the type of barcode). [...] The only disadvantage is that data must be coded in the barcode.
Oddthinking
  • 140,378
  • 46
  • 548
  • 638
  • Note that the error rate quoted here is that for mis-scanning (and I suspect accounts for the error correction intrinsic in the code). It leaves open the possibility of errors in a database backing the system (because most stores don't code the price in the barcode, but rather a product ID for use by the inventory control system). – dmckee --- ex-moderator kitten May 23 '11 at 17:41
  • 1
    @dmckee, very true, which is why I was sure to include that last sentence in the quote. For fairness, the corresponding error - wrongly printed price-tags - should also be included in the mix. – Oddthinking May 24 '11 at 00:11